980 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 10, NO.
3, MARCH 2011
Exact Performance of Two-Way AF Relaying in
Nakagami-𝑚 Fading Environment
Jing Yang, Pingzhi Fan, Trung Q. Duong, and Xianfu Lei
Abstract—The performance of two-way amplify-and-forward
(AF) relaying networks over independently but not necessarily
identically distributed (i.n.i.d.) Nakagami-𝑚 fading channels,
with integer and integer plus one-half values of fading parameter
𝑚, is studied. Closed-form expressions for the cumulative distri-
bution function (CDF), probability density function (PDF), and
moment generating function (MGF) of the end-to-end signal-
to-noise ratio (SNR) are presented. Utilizing these results, we
analyze the performance of two-way AF relaying system in
terms of outage probability, average symbol error rate (SER),
and average sum-rate. Simulations are performed to verify the
correctness of our theoretical analysis.
Index Terms—Two-way relaying network, outage probability,
average symbol error rate (SER), average sum-rate, Nakagami-𝑚
fading.
I. I NTRODUCTION
I N conventional cooperative communication networks,
when all terminals are operated in half-duplex fashion,
the transmission of one information symbol from the source
terminal to the destination terminal occupies two channel uses
[1]. This leads to a loss in spectral efficiency due to the pre-log
factor one-half in corresponding to sum-rate expressions. To
improve the spectral efficiency in half-duplex relaying, two-
way relaying was proposed in [1], [2], where two source nodes
simultaneously send their information to the relay node in
the multiple-access (MA) phase and the relay node broadcasts
the received signal to the two source nodes in the broadcast
(BC) phase. In [1], Rankov and Wittneben have introduced
the two-way relaying protocols, known as two-way amplify-
and-forward (AF) relaying and two-way decode-and-forward
(DF) relaying.
Recently, the performance of the conventional one-way co-
operation system has been well investigated [3]–[9]. However,
scanning the up-to-date open technical literature, the number
Manuscript received June 28, 2010; revised October 17, 2010; accepted
December 12, 2010. The associate editor coordinating the review of this paper
and approving it for publication was S. Ghassemzadeh.
This work was supported by the National 863 High-Tech R & D
Program (No.2009AA01Z238), the National Science Foundation of China
(No.61032002/60872013/60872014/60902023), the 111 Project (No.111-2-
14), the Sino-Swedish Cooperation Program (No.2008DFA12160), the Open
Research Fund of State Key Laboratory of Integrated Services Networks
(No.ISN11-04), and the Fundamental Research Funds for the Central Uni-
versities (No.2010XS21).
J. Yang, P. Fan, and X. Lei are with the Institute of Mobile Commu-
nications, Southwest Jiaotong University, Chengdu, Sichuan, 610031, P. R.
of China. J. Yang and X. Lei are also with The State Key Laboratory
of Integrated Services Networks, Xidian University, Xi’an, China (e-mail:
jingyang95@gmail.com, p.fan@ieee.org, xflei81@yahoo.com.cn).
T. Q. Duong is with the Blekinge Institute of Technology, SE-371 79
Karlskrona, Sweden (e-mail: quang.trung.duong@bth.se).
Digital Object Identifier 10.1109/TWC.2011.011111.101141
1536-1276/11$25.00 ⃝
c 2011 IEEE
of published works concerning performance analysis of two-
way relaying system is relatively small. Han et al. presented
an upper and lower bound for average sum-rate and an upper
bound for pairwise error probability (PEP) in [10]. Ping and
Ting investigated and compared the average rate performances
in one-way and two-way AF relaying network [11]. Qiang
et al. derived closed-form results for the outage probability
in [12]. Later, Duong et al. obtained the exact closed-form
expressions for outage probability, average symbol error rate
(SER), and average sum-rate in [13].
While the aforementioned works assumed Rayleigh fading
scenarios, to the best of our knowledge, the performance of
two-way relaying system over generalized Nakagami-𝑚 fading
has not been investigated. It is well known that Nakagami-
𝑚 fading spans via the 𝑚 parameter a wide range of fading
scenarios that is typical in realistic wireless relay applications,
which includes the one-sided Gaussian distribution (𝑚=0.5)
and Rayleigh fading (𝑚=1) as special cases. In practice, the
radio links between the two source terminals and the relay
may experience different fading scenario, which is a realistic
assumption for a scenario where two source terminals are
positioned different distance away from the relay node.
In this paper, we focus on the performance analysis of
two-way AF relaying system in generally independently but
not necessarily identically distributed (i.n.i.d.) Nakagami-𝑚
fading channels. Specifically, we derive exact expressions for
outage probability, average SER, and average sum-rate. We
first present tractable forms for the statistics of the end-
to-end signal-to-noise ratio (SNR), e.g., cumulative distribu-
tion function (CDF), probability density function (PDF), and
moment generating function (MGF). These new results are
then applied to study the end-to-end performance of two-way
relaying network, resulting in exact expressions for outage
probability, average SER and average sum-rate. Note that for
the case of Rayleigh fading channels, our general expressions
for Nakagami-𝑚 fading reduce to the previously published
results presented in [13] as special cases. Analytical formulas
are also verified by simulations.
II. S YSTEM AND C HANNEL M ODEL
A simple two-way relaying system model operating over
i.n.i.d. Nakagami-𝑚 fading channels is illustrated in Fig. 1.
Two source nodes T 1 and T 2 wish to exchange information
with each other via an AF relay node R. Each transmission
period is divided into two transmission phases: MA phase
and BC phase. In the MA phase, T 1 and T 2 simultaneously
send their information messages to R. In the BC phase, R
normalizes the received signal based on its transmit power
YANG et al.: EXACT PERFORMANCE OF TWO-WAY AF RELAYING IN NAKAGAMI-𝑀 FADING ENVIRONMENT 981

MA MA by

2𝑚𝑚 𝑖 −𝑏𝑖 𝛾 𝑚𝑖 −1 𝑚
(𝑚𝑗 − 1)! ∑ ∑ ∑ 1 (𝑝𝑗 )
𝑗 −1 𝑝𝑗
𝑖 𝑒
𝐹𝛾𝑇𝑖 (𝛾) = 1 −
T1 R T2 𝛾¯𝑖𝑚𝑖 Γ(𝑚𝑖 )Γ(𝑚𝑗 ) 𝑝 =0 𝑝 =0 𝑝𝑗 ! 𝑘
𝑖 𝑗 𝑘=0
h1 ( )( )− 𝜎𝑖2+1 ( )𝑝 + 𝜎𝑖 +1
h2 𝑚𝑖 − 1 𝑚𝑖 2𝑚𝑗 𝑗 2
×
𝑝𝑖 𝛾¯𝑖 𝛾¯𝑗
BC BC
× 𝛾 𝛿𝑖 𝐾𝜎𝑖 +1 (𝑎𝛾), (4)
Fig. 1. System model. where√ 𝛿𝑖 = 𝑚𝑖 + 𝑝𝑗 , 𝜎𝑖 = 𝑝𝑖 − 𝑘, 𝑏𝑖 = 𝑚𝑖 /¯ 𝛾𝑖 + 2𝑚𝑗 /¯
𝛾𝑗 ,
𝑎 = 2 2𝑚𝑖 𝑚𝑗 /(¯ 𝛾𝑖 𝛾¯𝑗 ) and 𝐾𝑣 (⋅) is the v-th order modified
Bessel function of the second kind [14, Eq. (8.432.6)].
constraint and broadcasts to T 1 and T 2 . As in [13], for Proof: See Appendix A.
simplicity of derivation, we assume equal transmit power Theorem 2: When 𝑚 is integer plus one half, the CDF of
(denoted as P𝑠 ) at T 1 , T 2 and R and equal variance (denoted 𝛾𝑇𝑖 is given by
as N 0 ) for the additive white Gaussian noise (AWGN) at the √ ( ) 𝑚2𝑖 − 14 ( ) 𝑚𝑗 − 1
three terminals. Moreover, we presume that T 1 and T 2 have the 𝜋 𝑚𝑖 2𝑚𝑗 2 4
𝐹𝛾𝑇𝑖 (𝛾) = 1 −
perfect knowledge of the corresponding channel coefficients Γ(𝑚𝑖 )Γ(𝑚𝑗 ) 𝛾¯𝑖 𝛾¯𝑗
so that the self-interference can be fully eliminated. We also ∑ ∑ 𝛾 2 𝑒 𝑖 𝐻𝛼−1 (𝑐𝑖 √𝛾)
𝑚𝑖 − 12 𝑚𝑗 − 12 𝛼 −𝑐2 𝛾
denote h1 and h2 as independent channel coefficients for × E, (5)
T1 → R and T2 → R links, respectively, which are modeled 𝑝 =0 𝑝 =0
2𝛼−1
𝑖 𝑗
as Nakagami-𝑚 random variables (RVs). Assume that the
channels are reciprocal, so the channel gains of T𝑖 → R where
2
∑
and R → T𝑖 (𝑖 = 1, 2) links are identical [1]. Hence, the 𝛼= (𝑚𝑙 − 𝑝𝑙 ) − 1,
instantaneous SNR of terminal T𝑖 in the BC phase, 𝛾𝑇𝑖 , can 𝑙=1
be given by [1], [10], [13] √ √
𝑚𝑖 2𝑚𝑗
𝛾𝑖 𝛾𝑗 𝑐𝑖 = + ,
𝛾𝑇𝑖 = , (1) 𝛾¯𝑖 𝛾¯𝑗
2𝛾𝑖 + 𝛾𝑗
( )− 𝑝2𝑖 ( )− 𝑝2𝑗
2𝑚𝑗
where 𝛾𝑖 = 𝑃𝑠 ∣ℎ𝑖 ∣2 /𝑁0 and 𝛾𝑗 = 𝑃𝑠 ∣ℎ𝑗 ∣2 /𝑁0 ( 𝑖, 𝑗 = 1, 2 (𝑚𝑖 + 𝑝𝑖 − 1
2 )!(𝑚𝑗 + 𝑝𝑗 − 12 )! 𝑚 𝛾
¯𝑖
𝑖
𝛾
¯𝑗
and 𝑖 ∕= 𝑗) are the instantaneous SNRs for T𝑖 → R and 𝐸= ,
4𝑝𝑖 +𝑝𝑗 𝑝𝑖 !𝑝𝑗 !(𝑚𝑖 − 𝑝𝑖 − 12 )!(𝑚𝑗 − 𝑝𝑗 − 12 )!
R → T𝑗 links, respectively. Note that { }𝛾0 = 𝑃𝑠 /𝑁 {0 is
} the
average transmit SNR and Ω𝑖 = 𝔼 ℎ2𝑖 , Ω𝑗 = 𝔼 ℎ2𝑗 are and
the respective variances of h𝑖 and h𝑗 , where 𝔼 {⋅} denotes the ⌊∑
2⌋
𝑛
𝑛−2𝑘
statistical average operator. Since h𝑖 is modeled as Nakagami- (−1)𝑘 (2𝑥)
𝐻𝑛 (𝑥) = 𝑛! , (6)
𝑚 RV, the instantaneous SNR, 𝛾i is a gamma distributed RV 𝑘! (𝑛 − 2𝑘)!
𝑘=0
with PDF given by
is the Hermite polynomials [14, Eq.(8.951].
𝑚𝑚 𝑖 𝑚 𝛾 Proof: See Appendix B.
− 𝑖
𝑓𝛾𝑖 (𝛾) = 𝑚𝑖
𝑖
𝛾 𝑚𝑖 −1 𝑒 𝛾¯𝑖 , (2) Corollary 1: When 𝑚 is integer plus one half, the PDF of
𝛾¯𝑖 Γ (𝑚𝑖 )
𝛾𝑇𝑖 is given by
where Γ(⋅) is the gamma function [14, Eq. (8.310.1)], 𝑚𝑖 is √ ( ) 𝑚2𝑖 − 14 ( ) 𝑚𝑗 − 1
𝜋 𝑚𝑖 2𝑚𝑗 2 4
the Nakagami fading parameter and 𝛾¯𝑖 = Ω𝑖 𝛾0 is the average 𝑓𝛾𝑇𝑖 (𝛾) =
SNR of the T𝑖 → R link. Hence, its CDF can be written as Γ(𝑚𝑖 )Γ(𝑚𝑗 ) 𝛾¯𝑖 𝛾¯𝑗
∑ ∑ 𝛾 2 𝑒 𝑖 𝐻𝛼+1 (𝑐𝑖 √𝛾)
𝑚𝑖 − 12 𝑚𝑗 − 12 𝛼 −1 −𝑐2 𝛾

Γ (𝑚𝑖 , (𝑚𝑖 /¯
𝛾𝑖 ) 𝛾) × E. (7)
𝐹𝛾𝑖 (𝛾) = 1 − , (3) 2𝛼+1
Γ (𝑚𝑖 ) 𝑝 =0 𝑝 =0
𝑖 𝑗

Proof: See Appendix C.

where Γ (⋅, ⋅) is the incomplete gamma function defined in
[14, Eq. (8.350.2)].
III. P ERFORMANCE A NALYSIS OF T WO -WAY AF
R ELAYING N ETWORK
We first present the CDF and PDF of 𝛾𝑇𝑖 , which will be
used for the derivation of the performance metrics.
Theorem 1: When 𝑚 is integer, the CDF of 𝛾𝑇𝑖 is given
A. Outage Probability
The outage probability 𝑃𝑜𝑢𝑡,𝑇𝑖 (𝛾𝑡ℎ ) at T 𝑖 is defined as the
probability that the instantaneous SNR falls below a given
threshold 𝛾𝑡ℎ ,
𝑃𝑜𝑢𝑡,𝑇𝑖 (𝛾𝑡ℎ ) = Pr (𝛾𝑇𝑖 < 𝛾𝑡ℎ ) = 𝐹𝛾𝑇𝑖 (𝛾𝑡ℎ ) , (8)
where 𝐹𝛾𝑇𝑖 (𝛾𝑡ℎ ) is the CDF of the instantaneous SNR at
terminal T 𝑖 , evaluated at 𝛾=𝛾𝑡ℎ .
982 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 10, NO. 3, MARCH 2011

√
It is clear now that by using (4), (5), and (8), the outage 𝜋(2𝑎)𝜎𝑖 +1 Γ(𝛿𝑖 + 𝜎𝑖 + 1)Γ(𝛿𝑖 − (𝜎𝑖 + 1))
𝑠2 = ( )
probability of two-way AF relaying network over Nakagami- (𝑏𝑖 + 𝑠 + 𝑎)𝛿𝑖 +𝜎𝑖 +1 Γ 𝛿𝑖 + 12
( )
𝑚 fading channels with integer and integer plus one half 3 1 𝑏𝑖 + 𝑠 − 𝑎
values of 𝑚 at terminal 𝑇𝑖 can be easily expressed in closed × 2 𝐹1 𝛿𝑖 + 𝜎𝑖 + 1, 𝜎𝑖 + ; 𝛿𝑖 + ; ,
2 2 𝑏𝑖 + 𝑠 + 𝑎
form.
For the case of Rayleigh fading channel, i.e., 𝑚 = 1, we
√
can see that our result particularizes to [13, Eq. (16)]. 𝜋(2𝑎)𝜎𝑖 Γ(𝛿𝑖 + 𝜎𝑖 + 1)Γ(𝛿𝑖 − 𝜎𝑖 + 1)
When 𝑚 = 0.5, the Nakagami distribution corresponds to a 𝑠3 = −
(𝑏𝑖 + 𝑠 + 𝑎)𝛿𝑖 +𝜎𝑖 +1 Γ(𝛿𝑖 + 32 )
one-sided Gaussian distribution, which is an important special ( )
1 3 𝑏𝑖 + 𝑠 − 𝑎
case as it represents the worst case of fading channels. Espe- × 2 𝐹1 𝛿𝑖 + 𝜎𝑖 + 1, 𝜎𝑖 + ; 𝛿𝑖 + ; ,
cially, the obtained result of outage probability for 𝑚 = 0.5 2 2 𝑏𝑖 + 𝑠 + 𝑎
will have a significant practical usefulness in wireless relaying and 𝑝 𝐹𝑞 (𝑎1 , ..., 𝑎𝑝 ; 𝑏1 , ..., 𝑏𝑞 ; 𝑥) is the generalized hypergeo-
systems when quality of service requirements mandate design- metric function [14, Eq.(9.14.1)].
ing for worst-case fading. For 𝑚 = 1, (11) exactly agrees with [13, Eq. (9)].
B. Average Symbol Error Rate 2) Integer plus one-half values of 𝑚: In this case, with the
help of (7), Φ𝛾𝑇𝑖 (𝑠) can be determined by
In order to obtain the average SER for a large variety
of modulation schemes, the MGF based approach [15] will √ ( ) 𝑚2𝑖 − 14 ( ) 𝑚𝑗 − 1
be used in this paper. Based on the fact that Φ𝛾𝑇𝑖 (𝑠) = 𝜋 𝑚𝑖 2𝑚𝑗 2 4
Φ𝛾𝑇𝑖 (𝑠) =
𝔼𝛾𝑇𝑖 {exp(−𝑠𝛾)}, to evaluate the average SER, the PDF Γ(𝑚𝑖 )Γ(𝑚𝑗 ) 𝛾¯𝑖 𝛾¯𝑗
𝑓𝛾𝑇𝑖 (𝛾) and the MGF Φ𝛾𝑇𝑖 (𝑠) will be firstly obtained in the 𝑚𝑖 − 12 𝑚𝑗 − 12 ⌊ 𝛼+1
2 ⌋
∑ ∑ ∑ (−1)𝑘 E × (𝛼 + 1)!(2𝑐𝑖 )𝛼+1−2𝑘
following. ×
1) Integer values of 𝑚: In this case, 𝑓𝛾𝑇𝑖 (𝛾) can be found 𝑝𝑖 =0 𝑝𝑗 =0 𝑘=0
2𝛼+1 𝑘!(𝛼 + 1 − 2𝑘)!
by taking the first derivative of (4) with respect to 𝛾, yielding ∫ ∞
2 2𝛼−2𝑘−1
× 𝑒−(𝑐𝑖 +𝑠)𝛾 𝛾 2 𝑑𝛾. (12)
2𝑚𝑚
𝑚∑𝑖 −1 𝑚
∑ ∑ 1 (𝑝𝑗 )
𝑗 −1 𝑝𝑗
𝑖 (𝑚𝑗 − 1)!
𝑖
0
𝑓𝛾𝑇𝑖 (𝛾) = − 𝑚𝑖
𝛾¯𝑖 Γ(𝑚𝑖 )Γ(𝑚𝑗 ) 𝑝 =0 𝑝 =0 𝑝𝑗 ! 𝑘 From [14, Eq. (3.326.2)], (12) can be calculated as
𝑖 𝑗 𝑘=0
( )( ) 𝜎 +1
− 𝑖2 ( ) 2𝑝𝑗 +𝜎𝑖 +1
√ ( ) 𝑚2𝑖 − 14 ( ) 𝑚𝑗 − 1
𝑚𝑖 − 1 𝑚𝑖 2𝑚𝑗 2
𝜋 𝑚𝑖 2𝑚𝑗 2 4
× Φ𝛾𝑇𝑖 (𝑠) =
𝑝𝑖 𝛾¯𝑖 𝛾¯𝑗 Γ(𝑚𝑖 )Γ(𝑚𝑗 ) 𝛾¯𝑖 𝛾¯𝑗
× [𝐿′1 (𝛾)𝐿2 (𝛾) + 𝐿1 (𝛾)𝐿′2 (𝛾)] , (9) 𝑚𝑖 − 12 𝑚𝑗 − 12 ⌊ 𝛼+1
2 ⌋
∑ ∑ ∑ (−1)𝑘 (𝛼 + 1)!(2𝑐𝑖 )𝛼−2𝑘+1
Recalling the relationship between Φ𝛾𝑇𝑖 (𝑠) and 𝑓𝛾𝑇𝑖 (𝛾), ×
Φ𝛾𝑇𝑖 (𝑠) can be determined by 𝑝 =0 𝑝 =0
2𝛼+1 𝑘!(𝛼 − 2𝑘 + 1)!
𝑖 𝑗 𝑘=0
−𝑢
2𝑚𝑚𝑖 (𝑚𝑗 − 1)! ∑
𝑚𝑖 −1 𝑚
∑ ∑ 1 (𝑝𝑗 )
𝑗 −1 𝑝𝑗 × (𝑐2𝑖 + 𝑠) Γ(𝑢)E, (13)
Φ𝛾𝑇𝑖 (𝑠) = − 𝑚𝑖 𝑖
𝛾¯𝑖 Γ(𝑚𝑖 )Γ(𝑚𝑗 ) 𝑝 =0 𝑝 =0 𝑝𝑗 ! 𝑘 where 𝑢 = (2𝛼 − 2𝑘 + 1)/2.
𝑖 𝑗 𝑘=0
( )( )− 𝜎𝑖2+1 ( ) 2𝑝𝑗 +𝜎 𝑖 +1 For 𝑚 = 0.5, one can obtain the MGF of 𝛾𝑇𝑖 from (13) in
𝑚𝑖 − 1 𝑚𝑖 2𝑚𝑗 2
one-sided Gaussian distribution fading channel.
×
𝑝𝑖 𝛾¯𝑖 𝛾¯𝑗 The derived MGF, Φ𝛾𝑇𝑖 (𝑠) in (11) and (13), can be used to
∫ ∞
× [𝐿′1 (𝛾)𝐿2 (𝛾) + 𝐿1 (𝛾)𝐿′2 (𝛾)] 𝑒−𝑠𝛾 𝑑𝛾, (10) evaluate the average SER. For instance, the average SER for
0 M-PSK at T𝑖 can be expressed as [15, Eq. (9.15)]
where 𝐿1 (𝛾) = 𝑒−𝑏𝑖 𝛾 , 𝐿2 (𝛾) = 𝛾 𝛿𝑖 𝐾𝜎𝑖 +1 (𝑎𝛾) and 𝐿′𝑖 (𝛾) ∫ ( )
1 (𝜋−𝜋/𝑀) 𝑔
with 𝑖 = 1, 2 is the first derivative of 𝐿𝑖 (𝛾). 𝑃𝑠𝑇𝑖 = Φ𝛾 𝑇 𝑖 𝑑𝜃, (14)
From [14, Eq. (8.486.12)] and [14, Eq. (6.621.3)], (10) can 𝜋 0 sin2 𝜃
be calculated in closed-form as
where 𝑔 = sin2 (𝜋/𝑀 ).
2𝑚𝑚
𝑚
∑ 𝑖 −1 𝑚
∑ ∑ 1 (𝑝𝑗 )
𝑗 −1 𝑝𝑗
𝑖 (𝑚𝑗 − 1)!
𝑖
The exact SER expression given in (14) readily enables us
Φ𝛾𝑇𝑖 (𝑠) = − 𝑚𝑖
𝛾¯𝑖 Γ(𝑚𝑖 )Γ(𝑚𝑗 ) 𝑝 =0 𝑝 =0 𝑝𝑗 ! 𝑘 to evaluate the SER performance by utilizing some efficient
𝑖 𝑗 𝑘=0
numerical integration techniques.
( )( )− 𝜎𝑖2+1 ( ) 2𝑝𝑗 +𝜎 𝑖 +1
𝑚𝑖 − 1 𝑚𝑖 2𝑚𝑗 2
×
𝑝𝑖 𝛾¯𝑖 𝛾¯𝑗
C. Average Sum-Rate
× [𝑏𝑖 𝑠1 + (𝛿𝑖 − (𝜎𝑖 + 1))𝑠2 + 𝑎𝑠3 ], (11)
The average sum-rate of the two-way relaying networks is
where
√ defined as the summation of average rate of T 1 and T 2 and
𝜋(2𝑎)𝜎𝑖 +1 Γ(𝛿𝑖 + 𝜎𝑖 + 2)Γ(𝛿𝑖 − 𝜎𝑖 ) can be shown as [13]
𝑠1 = −
(𝑏𝑖 + 𝑠 + 𝑎)𝛿𝑖 +𝜎𝑖 +2 Γ(𝛿𝑖 + 32 ) { } ∑
( ) 2
∑ 1
2
3 3 𝑏𝑖 + 𝑠 − 𝑎 𝑅sum = 𝔼𝛾𝑇𝑖 log2 (1 + 𝛾𝑇𝑖 ) = 𝑅sum,𝑇𝑖 . (15)
× 2 𝐹1 𝛿𝑖 + 𝜎𝑖 + 2, 𝜎𝑖 + ; 𝛿𝑖 + ; , 2
2 2 𝑏𝑖 + 𝑠 + 𝑎 𝑖=1 𝑖=1
YANG et al.: EXACT PERFORMANCE OF TWO-WAY AF RELAYING IN NAKAGAMI-𝑀 FADING ENVIRONMENT 983

1) Integer values of 𝑚: Based on (9), we can rewrite the 𝐻temp2 =

summand of (15) as ⎡  ⎤
1  (2, 1)
𝑏𝑖 −𝑎 
⎢  (1, 1) , (1, 1) ; ( 12 , 1) ⎥
∑ 1 (𝑝𝑗 ) ⎢  ⎥,
1,2,0,1,2
1 2𝑚𝑚
𝑚
∑ 𝑖 −1 𝑚
∑ 𝑗 −1 𝑝𝑗 𝐻1,[2:1],0,[2:2] 
𝑖 (𝑚𝑗 − 1)! ⎣ ⎦
𝑖

𝑅sum,𝑇𝑖 =−  −−
2 ln 2 𝛾¯𝑖𝑚𝑖 Γ(𝑚𝑖 )Γ(𝑚𝑗 ) 𝑝 =0 𝑝 =0 𝑝𝑗 ! 𝑘 2𝑎  (1, 1) , (0, 1) ; (𝜎𝑖 , 1), (−𝜎𝑖 , 1)
𝑖 𝑗 𝑘=0 𝑏𝑖 −𝑎
( )( )− 𝜎𝑖2+1 ( ) 2𝑝𝑗 +𝜎 𝑖 +1
𝑚𝑖 − 1 𝑚𝑖 2𝑚𝑗 2
× and 𝜉𝑖 = 𝜎𝑖 + 1.
𝑝𝑖 𝛾¯𝑖 𝛾¯𝑗
∫ ∞ The average sum-rate in two-way AF relaying network over
× [𝐿′1 (𝛾)𝐿2 (𝛾) + 𝐿1 (𝛾)𝐿′2 (𝛾)] ln(1 + 𝛾) 𝑑𝛾. (16) Nakagami-𝑚 fading channels with integer values of 𝑚 can be
0
evaluated by using (18). As can be seen from (18), the fading
After some elementary manipulation, (16) can be expressed parameter 𝑚 affects the average sum-rate performance.
as When 𝑚 = 1, our obtained average sum-rate reduces to
[13, Eq. (15)].
1 2𝑚𝑚
𝑚
∑ 𝑖 −1 𝑚
∑ ∑ 1 (𝑝𝑗 )
𝑗 −1 𝑝𝑗
2) Integer plus one-half values of 𝑚: Based on (7), we can
𝑖 (𝑚𝑗 − 1)!
𝑖

𝑅sum,𝑇𝑖 = − rewrite the summand of (15) as

2 ln 2 𝛾¯𝑖𝑚𝑖 Γ(𝑚𝑖 )Γ(𝑚𝑗 ) 𝑝 =0 𝑝 =0 𝑝𝑗 ! 𝑘
𝑖 𝑗 𝑘=0
( )( )− 𝜎𝑖2+1 ( )𝑝 + 𝜎𝑖 +1 𝑚𝑖 − 12 𝑚𝑗 − 12 ⌊ 𝛼+1
∑ ∑ ∑ 2 ⌋ √
𝑚𝑖 − 1 𝑚𝑖 2𝑚𝑗 𝑗 2 1 𝜋
× 𝑅sum,𝑇𝑖 =
𝑝𝑖 𝛾¯𝑖 𝛾¯𝑗 2 ln 2 𝑝 =0 𝑝 =0 Γ(𝑚𝑖 )Γ(𝑚𝑗 )
𝑖 𝑗 𝑘=0
× (𝑡1 + 𝑡2 + 𝑡3 ), (17) ( ) 𝑚2𝑖 − 14 ( ) 𝑚2𝑗 − 14
𝑚𝑖 2𝑚𝑗 (−1)𝑘 (𝛼 + 1)!(2𝑐𝑖 )𝛼+1−2𝑘 E
×
where 𝛾¯𝑖 𝛾¯𝑗 2𝛼+1 𝑘!(𝛼 + 1 − 2𝑘)!
∫ ∞
∫ ∞ ×
2 2𝛼−2𝑘−1
𝑒−𝑐𝑖 𝛾 𝛾 2 ln(1 + 𝛾) 𝑑𝛾. (19)
𝑡1 = −𝑏𝑖 𝛾 𝛿𝑖 𝑒−𝑏𝑖 𝛾 ln (1 + 𝛾) 𝐾𝜎𝑖 +1 (𝑎𝛾) 𝑑𝛾, 0
0
Finally, using [14, Eq.(4.352.1)] yields the closed-form ex-
𝑡2 = (𝛿𝑖 − 𝜎𝑖 − 1)
∫ ∞ pression of the summand 𝑅sum,𝑇𝑖 which helps us to obtain
× 𝛾 𝛿𝑖 −1 𝑒−𝑏𝑖 𝛾 ln (1 + 𝛾) 𝐾𝜎𝑖 +1 (𝑎𝛾) 𝑑𝛾, the average sum-rate as follows:
∫ ∞ 0 1 1 𝛼+1
2 𝑚𝑖 − 𝑚𝑗 − ⌊ 2 ⌋ √
𝑡3 = −𝑎 𝛾 𝛿𝑖 𝑒−𝑏𝑖 𝛾 ln (1 + 𝛾) 𝐾𝜎𝑖 (𝑎𝛾) 𝑑𝛾. 1 ∑ ∑2 ∑2 ∑ 𝜋
𝑅sum =
0 2 ln 2 𝑖=1 𝑝 =0 𝑝 =0 Γ(𝑚𝑖 )Γ(𝑚𝑗 )
𝑖 𝑗 𝑘=0
( ) 𝑚2𝑖 − 14 ( ) 𝑚2𝑗 − 14
∫ ∞to evaluate 𝑅sum,𝑇𝑖 , we will utilize the result of
In order 𝑚𝑖 2𝑚𝑗 (−1)𝑘 (𝛼 + 1)!(2𝑐𝑖 )𝛼+1−2𝑘 E
𝐽 = 0 𝑥𝑚 𝑒−𝑝𝑥 ln(1 + 𝑥)𝐾𝑛 (𝑞𝑥)𝑑𝑥, which is derived in ×
𝛾¯𝑖 𝛾¯𝑗 2𝛼+1 𝑘!(𝛼 + 1 − 2𝑘)!
Appendix D. Applying (37) into (17) yields the closed-form
expression of the summand 𝑅sum,𝑇𝑖 which helps us to obtain × 𝑅T𝑖 ,𝐷 , (20)
the average sum-rate 𝑅sum as follows: where
2 𝑚𝑖 −1 𝑚 𝑗 −1 𝑝𝑗 + 1; 𝑐2𝑖 )
1 𝐹1 (𝑢; 𝑢
1 ∑ ∑ ∑ ∑ 2𝑚𝑚𝑖 (𝑚𝑗 − 1)! 1
𝑖 𝑅𝑇𝑖 ,𝐷 = 𝜋 − 𝑐−2𝑢
𝑖 Γ(𝑢)
𝑅sum =− 𝑢 sin(𝑢𝜋)
2 ln 2 𝑖=1 𝑝 =0 𝑝 =0 𝛾¯𝑖𝑚𝑖 Γ(𝑚𝑖 )Γ(𝑚𝑗 ) 𝑝𝑗 ! [ 2
]
2 𝐹2 (1, 1; 2, 2 − 𝑢; 𝑐𝑖 ) 2
𝑖 𝑗 𝑘=0
( )( )( )− 𝜎𝑖2+1 ( )𝑝 + 𝜎𝑖 +1 × 2 ln 𝑐𝑖 − 𝜓(𝑢) + 𝑐𝑖 , (21)
𝑝𝑗 𝑚𝑖 − 1 𝑚𝑖 2𝑚𝑗 𝑗 2 1−𝑢
×
𝑘 𝑝𝑖 𝛾¯𝑖 𝛾¯𝑗 and 𝜓(⋅) denotes the Euler psi function [14, Eq.(8.360.1)],
× (𝑅𝑇𝑖 ,𝐴 + 𝑅𝑇𝑖 ,𝐵 + 𝑅𝑇𝑖 ,𝐶 ), (18) j = 1, 2 and j ∕= i.
Again, it can also be seen from (20), average sum-rate
where performance depends on the fading severity parameter 𝑚.
√ When 𝑚 = 0.5, our result given in (20) becomes the average
𝑅𝑇𝑖 ,𝐴 = − 𝑏𝑖 𝜋 (𝑏𝑖 − 𝑎)−𝛿𝑖 −1 × 𝐻temp1 , sum-rate of two-way AF relaying network over one-sided
√ −𝛿 Gaussian distribution fading channel.
𝑅𝑇𝑖 ,𝐵 = (𝛿𝑖 − 𝜎𝑖 − 1) 𝜋 (𝑏𝑖 − 𝑎) 𝑖 × 𝐻temp1 ,
√ −𝛿 −1
𝑅𝑇𝑖 ,𝐶 = − 𝑎 𝜋 (𝑏𝑖 − 𝑎) 𝑖 × 𝐻temp2 ,
IV. N UMERICAL R ESULTS
In this section, simulation is provided to demonstrate the
𝐻temp1 = validity and usefulness of our analytical expressions. For
⎡  ⎤ numerical results, the special function in the formulas can
1  (2, 1)
𝑏𝑖 −𝑎  be easily evaluated by using common mathematical software
⎢  (1, 1) , (1, 1) ; ( 12 , 1) ⎥
1,2,0,1,2
𝐻1,[2:1],0,[2:2] ⎢  ⎥, packages such as Mathematica or Maple. In order to compare
⎣  −− ⎦
 the results in [13], we also consider two examples: 1) i.i.d. case
2𝑎  (1, 1) , (0, 1) ; (𝜉𝑖 , 1), (−𝜉𝑖 , 1)
𝑏𝑖 −𝑎 (e.g., Ω1 = Ω2 = 1) and 2) i.n.i.d. case (e.g., Ω1 = 3, Ω2 = 5).
984 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 10, NO. 3, MARCH 2011

0 0
10 10

−1 −1
10 10

Average Symbol Error Rate

−2 −2
10 10
Outage Probability

−3 −3
10 Ω =1,Ω =1,m =1,m =1 10 Ω =1,Ω =1,m =1,m =1
1 2 1 2 1 2 1 2
Ω1=3,Ω2=5,m1=1,m2=1 Ω1=3,Ω2=5,m1=1,m2=1
−4 −4
10 Ω1=1,Ω2=1,m1=1,m2=2 10 Ω1=1,Ω2=1,m1=1,m2=2
Ω1=3,Ω2=5,m1=1,m2=2 Ω1=3,Ω2=5,m1=1,m2=2
−5
10 Ω1=1,Ω2=1,m1=2,m2=3 −5
10 Ω1=1,Ω2=1,m1=2,m2=3
Ω =3,Ω =5,m =2,m =3 Ω =3,Ω =5,m =2,m =3
1 2 1 2 1 2 1 2

−6 Theoretical −6 Theoretical
10 10
0 5 10 15 20 25 30 0 5 10 15 20 25 30
γ (dB) γ (dB)
0 0

Fig. 2. Outage probability for the two-way AF relaying at terminal T 2 with Fig. 4. Average SER for the two-way AF relaying at terminal T 2 with
integer values of m. integer values of m.
0
10 0
10

−1
−1
10 10

Average Symbol Error Rate

Outage Probability

−2
10
−2
10
Ω =1,Ω =1,m =0.5,m =0.5
1 2 1 2
Ω1=1,Ω2=1,m1=0.5,m2=0.5
Ω =3,Ω =5,m =0.5,m =0.5
1 2 1 2 −3
10 Ω1=3,Ω2=5,m1=0.5,m2=0.5
Ω =1,Ω =1,m =0.5,m =1.5
1 2 1 2
−3 Ω1=1,Ω2=1,m1=0.5,m2=1.5
10 Ω =3,Ω =5,m =0.5,m =1.5
1 2 1 2
Ω1=3,Ω2=5,m1=0.5,m2=1.5
Ω =1,Ω =1,m =1.5,m =2.5 −4
10
1 2 1 2
Ω1=1,Ω2=1,m1=1.5,m2=2.5
Ω =3,Ω =5,m =1.5,m =2.5
1 2 1 2
Ω1=3,Ω2=5,m1=1.5,m2=2.5
−4 Theoretical
10 −5 Theoretical
0 5 10 15 20 25 30 10
γ0 (dB) 0 5 10 15 20 25 30
γ0 (dB)

Fig. 3. Outage probability for the two-way AF relaying at terminal T 2 with

integer plus one-half values of m. Fig. 5. Average SER for the two-way AF relaying at terminal T 2 with
integer plus one-half values of m.

Different fading conditions, such as different fading parame-

ters, are considered. The Rayleigh faded performance is also
plotted. The solid lines in Figs. 2-7 show the theoretical results
for outage probability, average SER and average sum-rate,
which are obtained by the analytical expressions in this paper,
while the other lines represent the simulation results.
Fig. 2 and Fig. 3 draw respectively the outage probabil-
ity versus 𝛾0 with 𝛾𝑡ℎ = 3 at terminal 𝑇2 when fading
parameter is integer and integer plus one-half. As can be
clearly seen from both figures, analytical and simulated outage
probability curves match excellently. As expected, improved
fading severity conditions or fading conditions in two links
can significantly lower outage probability. For example, when
an outage probability equals to 10−2 , when fading parameters
change from 𝑚1 = 1, 𝑚2 = 1 to 𝑚1 = 2, 𝑚2 = 3, an SNR
gain of 15 dB can be achieved (for i.i.d case, see Fig. 2).
Fig. 4 and Fig. 5 present the average SER results for
QPSK modulation at terminal 𝑇2 , respectively for fading
parameter being integer and integer plus one-half. We see that
the simulation results also closely match with the analytical
ones. Results show that the average SER performance can
be significantly improved for various fading parameters. For
example, when average SER equals to 10−1 , when fading
parameters change from 𝑚1 = 0.5, 𝑚2 = 0.5 to 𝑚1 = 1.5,
𝑚2 = 2.5, an SNR gain of 10 dB can be achieved (for i.n.i.d
case, see Fig. 5).
Fig. 6 and Fig. 7 show the average sum-rate performance. It
can also be seen that the simulation results closely match with
the analytical ones. As expected, improved fading parameter in
two links can increase average sum-rate. For instance, when
average sum-rate equals to 2, when fading parameter from
𝑚1 = 0.5, 𝑚2 = 0.5 to 𝑚1 = 1.5, 𝑚2 = 2.5, a SNR gain of
5 dB can be achieved (for i.n.i.d case, see Fig. 7).
Note that for Rayleigh fading channels, all results confirm
with the counterparts in [13].
V. C ONCLUSIONS
In this paper, we have derived exact expressions of outage
probability, average SER, and average sum-rate for two-
way AF relaying networks under i.n.i.d. Nakagami-𝑚 fading
channels with integer and integer plus one-half values of
fading parameters. We have shown that our results reduce to
the previously published work for Rayleigh fading channels
as special cases. Our analysis has been verified by simulation.
YANG et al.: EXACT PERFORMANCE OF TWO-WAY AF RELAYING IN NAKAGAMI-𝑀 FADING ENVIRONMENT 985

12 11
Ω1=1,Ω2=1,m1=1,m2=1 Ω1=1,Ω2=1,m1=0.5,m2=0.5
10
Ω =3,Ω =5,m =1,m =1 Ω =3,Ω =5,m =0.5,m =0.5
1 2 1 2 1 2 1 2
10 9
Ω1=1,Ω2=1,m1=1,m2=2 Ω =1,Ω =1,m =0.5,m =1.5
1 2 1 2
Ω1=3,Ω2=5,m1=1,m2=2 8 Ω =3,Ω =5,m =0.5,m =1.5
1 2 1 2
8 Ω1=1,Ω2=1,m1=2,m2=3 Ω1=1,Ω2=1,m1=1.5,m2=2.5
Average Sum−Rate

Average Sum−Rate
7
Ω1=3,Ω2=5,m1=2,m2=3 Ω =3,Ω =5,m =1.5,m =2.5
6 1 2 1 2
6 Theoretical Theoretical
5

4
4
3

2 2

0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
γ (dB) γ (dB)
0 0

Fig. 6. Average sum-rate for the two-way AF relaying with integer values Fig. 7. Average sum-rate for the two-way AF relaying with integer plus
of m. one-half values of m.

A PPENDIX Using (26) and (27), after a simple transformation of the

A. Proof of Theorem 1 variables, 𝐼3 (𝛾) can be given by
By using (8), the CDF of 𝛾𝑇𝑖 , 𝐹𝛾𝑇𝑖 (𝛾) can be expressed as 𝑐−1 ∑
∑ 𝑘 ( )
(2ℓ)𝑘 𝑘
∫ ∞ [ ] 𝐼3 (𝛾) = (𝑐 − 1)!𝑒−2ℓ
𝛾𝑖 𝛾𝑗 𝑘! 𝑣
𝐹𝛾𝑇𝑖 (𝛾) = Pr ≤ 𝛾 ∣𝛾𝑖 𝑓𝛾𝑖 (𝛾𝑖 ) 𝑑𝛾𝑖 ∫ ( 𝑘=0
𝑣=0
)( )𝑣
2𝛾𝑖 + 𝛾𝑗 ∞
2ℓ𝑦 𝑦
∫0 𝛾 [ ] × 𝑥𝑎−1 exp −𝑏𝑥 − 𝑑𝑥. (28)
2𝛾𝛾𝑖 𝑥−𝑦 𝑥−𝑦
= Pr 𝛾𝑗 ≥ ∣𝛾𝑖 𝑓𝛾𝑖 (𝛾𝑖 ) 𝑑𝛾𝑖 𝑦
0 𝛾𝑖 − 𝛾
∫ ∞ [ ] The inner integral of (28) can be solved by using [14,
2𝛾𝛾𝑖
+ Pr 𝛾𝑗 ≤ ∣𝛾𝑖 𝑓𝛾𝑖 (𝛾𝑖 ) 𝑑𝛾𝑖 Eq. (3.471.9)], resulting in
𝛾 𝛾𝑖 − 𝛾
𝑐−1 ∑
∑ 𝑘 ∑
𝑎−1 ( )( )
= 𝐼1 (𝛾) + 𝐼2 (𝛾), (22) −2ℓ−𝑏𝑦 (2ℓ)𝑘 𝑘 𝑎−1
𝐼3 (𝛾) = (𝑐 − 1)!𝑒
where 𝑘! 𝑣 𝑟
𝑘=0 𝑣=0 𝑟=0
∫ 𝛾 ( ) 𝑟−𝑣+1
2ℓ𝑦 2 √
𝐼1 (𝛾) = 𝑓𝛾𝑖 (𝛾𝑖 ) 𝑑𝛾𝑖 × 𝑦 𝑎−𝑟+𝑣−1 2 𝐾𝑣−𝑟−1 (2 2ℓ𝑦𝑏). (29)
0 𝑏
Γ (𝑚𝑖 , (𝑚𝑖 /¯
𝛾𝑖 ) 𝛾)
= 𝐹𝛾𝑖 (𝛾) = 1 − , (23) With the help of (22), (23), (24) and (29), we get (4).
Γ (𝑚𝑖 )
∫ ∞ ) (
2𝛾𝛾𝑖 B. Proof of Theorem 2
𝐼2 (𝛾) = 𝐹𝛾𝑗𝑓𝛾𝑖 (𝛾𝑖 ) 𝑑𝛾𝑖
𝛾 𝛾𝑖 − 𝛾 We can rewrite (1) as follows:
Γ (𝑚𝑖 , (𝑚𝑖 /¯
𝛾𝑖 ) 𝛾) 𝑚𝑚
𝑖
𝑖

= − 𝑚𝑖 𝐼3 (𝛾), (24) 1
Γ (𝑚𝑖 ) 𝛾¯𝑖 Γ(𝑚𝑖 )Γ(𝑚𝑗 ) 𝛾𝑇𝑖 = , (30)
𝑋𝑖 + 𝑋𝑗
and
∫ ∞ ( ) where 𝑋𝑖 = 1/𝛾𝑖 , 𝑋𝑗 = 2/𝛾𝑗 .
−
𝑚𝑖 𝛾 𝑖 2𝛾𝑖 𝑚𝑗 𝛾
𝐼3 (𝛾) = 𝛾𝑖𝑚𝑖 −1 𝑒 𝛾
¯𝑖
Γ 𝑚𝑗 , 𝑑𝛾𝑖 . (25)
𝛾 𝛾𝑖 − 𝛾 𝛾¯𝑗 ∫∞
𝑠
Φ𝑋𝑖 (𝑠) = 𝔼𝑋𝑖 {exp (−𝑠𝑋𝑖 )} = 𝑒− 𝛾 𝑓𝛾𝑖 (𝛾) 𝑑𝛾. (31)
The integral 𝐼3 (𝛾) in (25) can be written in its general form
as 0
∫ ∞ ( )
𝑎−1 −𝑏𝑥 2ℓ𝑦 Substituting (2) into (31) and using [14, Eq.(3.471.9)], we get
𝐼3 (𝛾) = 𝑥 𝑒 Γ 𝑐, 2ℓ + 𝑑𝑥, (26)
𝑦 𝑥−𝑦 ( ) 𝑚𝑖 ( √ )
2 𝑚𝑖 𝑠 2 𝑚𝑖 𝑠
where 𝑎, 𝑐 are positive integers and 𝑏, ℓ are real numbers. By Φ𝑋𝑖 (𝑠) = 𝐾 𝑚𝑖 2 . (32)
Γ(𝑚𝑖 ) 𝛾¯𝑖 𝛾¯𝑖
using [14, Eq. (8.352.2)] and [14, Eq. (1.111)], the incomplete
gamma function of the integral in (26) can be expressed as Similarly, we get
( ) ( ) ( ) 𝑚2𝑗 ( √ )
2ℓ𝑦 2ℓ𝑦 2 2𝑚𝑗 𝑠 2𝑚𝑗 𝑠
Γ 𝑐, 2ℓ + = (𝑐 − 1)! exp −2ℓ − Φ𝑋𝑗 (𝑠) = 𝐾 𝑚𝑗 2 . (33)
𝑥−𝑦 𝑥−𝑦 Γ (𝑚𝑗 ) 𝛾¯𝑗 𝛾¯𝑗
𝑐−1 ∑
∑ 𝑘 𝑘
( )( )𝑣
(2ℓ) 𝑘 𝑦
× . (27) Let us define 𝑌𝑖 = 𝑋𝑖 + 𝑋𝑗 . Since 𝑋𝑖 and 𝑋𝑗 are statistically
𝑘! 𝑣 𝑥−𝑦
𝑣=0 𝑘=0 independent, by using (32) and (33), the MGF of 𝑌𝑖 can be
986 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 10, NO. 3, MARCH 2011

expressed as [5] T. A. Tsiftsis, G. K. Karagiannidis, P. T. Mathiopoulos, and S. A. Kot-

sopoulos, “Nonregenerative dual-hop cooperative links with selection
Φ𝑌𝑖 (𝑠) = Φ𝑋𝑖 (𝑠)Φ𝑋𝑗 (𝑠) diversity," EURASIP J. Wireless Commun. Netw., vol. 2006, pp. 1-8,
May 2006.
( ) 𝑚𝑖 ( ) 𝑚𝑗 [6] H. A. Suraweera, R. H. Y. Louie, Y. H. Li, G. K. Karagiannidis,
4 𝑚𝑖 𝑠 2 2𝑚𝑗 𝑠 2
= and B. Vucetic, “Two hop amplify-and-forward transmission in mixed
Γ(𝑚𝑖 )Γ(𝑚𝑗 ) 𝛾¯𝑖 𝛾¯𝑗 Rayleigh and Rician fading channels," IEEE Commun. Lett., vol. 13,
( √ ) ( √ ) no. 4, pp. 227-229, 2009.
𝑚𝑖 𝑠 2𝑚𝑗 𝑠 [7] X. Lei, P. Fan, L. Fan, K. K. Wong, E. G. Larsson, and T. Q. Duong,
× 𝐾 𝑚𝑖 2 𝐾 𝑚𝑗 2 . (34) “Closed-form performance analysis of multihop amplify-and-forward
𝛾¯𝑖 𝛾¯𝑗
transmission in Nakagami-𝑚 fading channels," to be submitted to IEEE
Signal Process. Lett..
Then, the CDF of 𝛾𝑇𝑖 , 𝐹𝛾𝑇𝑖 (𝛾), can be shown as [8] N. Yang, M. Elkashlan, and J. Yuan, “Outage probability of multiuser
relay networks in Nakagami-𝑚 fading channels," IEEE Trans. Veh.
𝐹𝛾𝑇𝑖 (𝛾) = 1 − 𝐹𝑌𝑖 (1/𝛾) = 1 − ℒ−1 (Φ𝑌𝑖 (𝑠)/𝑠)∣1/𝛾 , (35) Technol., vol. 59, no. 5, pp. 2120-2132, June 2010.
[9] J. Yang, P. Fan, Q. C. Chen, and X. Lei, “Delay-limited throughput
where ℒ−1 (⋅) stands for the inverse Laplace transform. Based analysis of dual-hop amplify-and-forward HARQ transmissions over
asymmetric fading channels," submitted to IEEE Trans. Veh. Technol.
on (34) and (35) and using [14, Eq.(8.468)], [16, vol.5, [10] Y. Han, S. H. Ting, C. K. Ho, and W. H. Chin, “Performance bounds
Eq.(2.2.1.11)], we obtain (5). for two-way amplify-and-forward relaying," IEEE Trans. Wireless Com-
mun., vol. 8, no. 1, pp. 432-439, Jan. 2009.
[11] J. Ping and S. H. Ting, “Rate performance of AF two-way relaying in
C. Proof of Corollary 1 low SNR region," IEEE Commun. Lett., vol. 13, no. 4, pp. 233-235,
Apr. 2009.
Utilizing the fact that the PDF of 𝑌𝑖 is 𝑓𝑌𝑖 (𝛾) = [12] Q. Li, S. H. Ting, A. Pandharipande, and Y. Han, “Adaptive two-way
ℒ−1 {Φ𝑌𝑖 (𝑠)}. Next, applying [16, vol.5, Eq.(2.2.1.11)], [14, relaying and outage analysis," IEEE Trans. Wireless Commun., vol. 8,
no. 6, pp. 3288-3299, June 2009.
Eq.(8.950.2)] and the identity 𝑓𝛾𝑇𝑖 (𝛾) = 𝛾 −2 𝑓𝑌𝑖 (1/𝛾), we [13] T. Q. Duong, L.-N. Hoang, and V. N. Q. Bao, “On the performance
get (7), which completes the proof. of two-way amplify-and-forward relay," IEICE Trans. Commun., vol.
E92-B, no. 12, pp. 3957-3959, Dec. 2009.
[14] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and
D.
∫ ∞Calculation of the Integral 𝐽 = Products, 6th edition. Academic Press, 2000.
[15] M. K. Simon and M. S. Alouini, Digital Communication over Fading
0
𝑥𝑚 𝑒−𝑝𝑥 ln(1 + 𝑥)𝐾𝑛 (𝑞𝑥)𝑑𝑥 Channels: A Unified Approach to Performance Analysis, 6th edition.
Using [16, vol. 3, Eq. (8.4.23.3)], [16, vol. 3, Eq. (8.4.6.5)], Wiley-Interscience, 2000.
[16] A. P. Prudnikov, Y. A. Brychkov, and O. I. Marichev, Integrals and
[16, vol. 3, Eq. (8.3.2.21)] and after some elementary manip- Series. Gordon and Breach Science, 1986, vol. 3,5.
ulations, we can write 𝐽 as [17] A. M. Mathai and R. K. Saxena, The H-Function with Applications in
∫ ∞{ [  ] Statistics and Other Disciplines. Wiley, 1978.
√  (1, 1) (1, 1)
𝐽= 𝜋 𝑚 (𝑞−𝑝)𝑥 1,2
𝑥 𝑒 𝐻2,2 𝑥  
(1, 1) (0, 1) Jing Yang received the B.S. degree in communica-
0 tions engineering from Sichuan Normal University,
[  ]} Chengdu, China, in 2006. She is currently working
 ( 1
, 1)
× 𝐻1,2 2𝑞𝑥 
2,0 2 𝑑𝑥, (36) toward her Ph.D. degree at the Institute of Mo-
(𝑛, 1), (−𝑛, 1) bile Communications, Southwest Jiaotong Univer-
sity, Chengdu, China. Her research interests include
𝐴,𝐵 the cross-layer optimization and cooperative diver-
where 𝐻𝐶,𝐷 [⋅] is the Fox’s 𝐻-function [16, vol. 3, sity.
Eq. (8.3.1.1)].
From [17, Eq. (2.6.2)], we get
√ −𝑚−1 Pingzhi Fan (M’93-SM’99) received his M.S. de-
𝐽 = 𝜋 (𝑝 − 𝑞)
⎡ 1  ⎤ gree in Computer Science from the Southwest Jiao-

𝑝−𝑞  (2, 1) tong University, PRC, in 1987, and Ph.D. degree
⎢  ⎥ in Electronic Engineering from the Hull University,
1,2,0,1,2 ⎢  (1, 1) , (1, 1) ; (0.5, 1) ⎥,
× 𝐻1,[2:1],0,[2:2] ⎣  ⎦
U.K., in 1994. He is currently a professor and
 −− director of the institute of mobile communications,
2𝑞 
𝑝−𝑞
(1, 1) , (0, 1) ; (𝑛, 1), (−𝑛, 1) Southwest Jiaotong University, PRC, and a guest
professor of Leeds University, UK (1997-), a guest
(37) professor Shanghai Jiaotong University (1999-). He
was a recipient of the UK ORS Award (1992),
′ ′
𝐾,𝑁,𝑁 ,𝑀,𝑀 and the NSFC Outstanding Young Scientist Award
where 𝐻𝐸,[𝐴:𝐶],𝐹,[𝐵:𝐷] [⋅] is the generalized Fox’s 𝐻-function (1998). He is the inventor of 22 patents, and the author of over 350 research
[17, Eq. (2.2.1)]. papers and eight books, including six books published by John Wiley & Sons
Ltd, RSP (1996), IEEE Press (2003,2006) and Springer (2004) and Nova
Science (2007), respectively. His research interests include spread-spectrum
R EFERENCES and CDMA technology, information theory & coding, sequence design &
applications, radio resource management, etc.
[1] B. Rankov and A. Wittneben, “Spectral efficient protocols for half-
duplex fading relay channels," IEEE J. Sel. Areas Commun., vol. 25,
no. 2, pp. 379-389, Feb. 2007.
[2] P. Larsson and N. Johansson, “Interference cancellation in wireless
relaying networks," United States Patent no. 7.336.930, Apr. 2004.
[3] J. N. Laneman, D. Tse, and G. Wornell, “Cooperative diversity in
wireless networks: efficient protocols and outage behavior," IEEE Trans.
Inf. Theory, vol. 50, no. 12, pp. 3062-3080, Dec. 2004.
[4] R. U. Nabar, H. Bölcskei, and F. W. Kneubuhler, “Fading relay channels:
performance limits and space-time signal design," IEEE J. Sel. Areas
Commun., vol. 22, no. 6, pp. 1099-1109, Aug. 2004.
YANG et al.: EXACT PERFORMANCE OF TWO-WAY AF RELAYING IN NAKAGAMI-𝑀 FADING ENVIRONMENT
Trung Q. Duong was born in HoiAn town, Quang
Nam province, Vietnam, in 1979. He received the
B.S. degree in electrical engineering from Ho Chi
Minh City University of Technology, Vietnam in
2002, and the M.Sc. degree in computer engineering
from Kyung Hee University, South Korea, in 2005.
In April 2004, he joined the faculty of electri-
cal engineering of Ho Chi Minh City University
of Transport, Vietnam. He was a recipient of the
Korean Government IT Scholarship Program for
International Graduate Students from 2003 to 2007.
He was awarded the Best Paper Award of IEEE Student Paper Contest -
IEEE Seoul Section in December 2006 and was listed in the finalist for the
best paper award at the IEEE Radio and Wireless Symposium, San Diego,
U.S.A. in 2009. He has been acting as an active reviewer for numerous
journals/conferences and a TPC member for more than 10 conferences. He has
been a TPC co-chair of the International Workshop on Multimedia Streaming
(MultiStreams2011) and a visiting scholar at Polytechnic Institute of New
York University from December 2009 to January 2010. In December 2007,
he joined the Radio Communication Group, Blekinge Institute of Technology,
987
Sweden as a research staff working toward his Ph.D. degree in radio
communications. He is the author or coauthor of some 50 technical papers in
the area of wireless and mobile communications. His current research interests
include cross-layer optimization, cooperative communications, cognitive radio
networks, and applied game theory in wireless systems.
Xianfu Lei is with Institute of Mobile Communi-
cations (IMC), Key Lab of Information Coding and
Transmission, Southwest Jiaotong University, China.
He is serving as the treasurer of IEEE Information
Theory Society Chengdu Chapter. He served as
workshop co-chair of PerMedia’10. He serves as
publication chair for IWSDA’11 and TPC member
for Globecom’11, ICC’11, WCNC’11, VTC’11, etc.
His research interests include digital communica-
tions over fading channels, relay-assisted and coop-
erative communications, cognitive communications,
wireless communication theory, etc.